Anomalies in Vote Counts and Their Effects on Election 2020
A Quantitative Analysis of Decisive Vote Updates in Michigan, Wisconsin, and Georgia on and after Election Night.
In the early hours of November 4th, 2020, Democratic candidate Joe Biden received several major “vote spikes” that substantially — and decisively — improved his electoral position in Michigan, Wisconsin, and Georgia. Much skepticism and uncertainty surrounds these “vote spikes.” Critics point to suspicious vote counting practices, extreme differences between the two major candidates’ vote counts, and the timing of the vote updates, among other factors, to cast doubt on the legitimacy of some of these spikes. While data analysis cannot on its own demonstrate fraud or systemic issues, it can point us to statistically anomalous cases that invite further scrutiny.
This is one such case: Our analysis finds that a few key vote updates in competitive states were unusually large in size and had an unusually high Biden-to-Trump ratio. We demonstrate the results differ enough from expected results to be cause for concern.
With this report, we rely only on publicly available data from the New York Times to identify and analyze statistical anomalies in key states. Looking at 8,954 individual vote updates (differences in vote totals for each candidate between successive changes to the running vote totals, colloquially also referred to as “dumps” or “batches”), we discover a remarkably consistent mathematical property: there is a clear inverse relationship between difference in candidates’ vote counts and and the ratio of the vote counts. (In other words, it's not surprising to see vote updates with large margins, and it's not surprising to see vote updates with very large ratios of support between the candidates, but it is surprising to see vote updates which are both).
The significance of this property will be further explained in later sections of this report. Nearly every vote update, across states of all sizes and political leanings follow this statistical pattern. A very small number, however, are especially aberrant. Of the seven vote updates which follow the pattern the least, four individual vote updates — two in Michigan, one in Wisconsin, and one in Georgia — were particularly anomalous and influential with respect to this property and all occurred within the same five hour window.
In particular, we are able to quantify the extent of compliance with this property and discover that, of the 8,954 vote updates used in the analysis, these four decisive updates were the 1st, 2nd, 4th, and 7th most anomalous updates in the entire data set. Not only does each of these vote updates not follow the generally observed pattern, but the anomalous behavior of these updates is particularly extreme. That is, these vote updates are outliers of the outliers.
The four vote updates in question are:
An update in Michigan listed as of 6:31AM Eastern Time on November 4th, 2020, which shows 141,258 votes for Joe Biden and 5,968 votes for Donald Trump
An update in Wisconsin listed as 3:42AM Central Time on November 4th, 2020, which shows 143,379 votes for Joe Biden and 25,163 votes for Donald Trump
A vote update in Georgia listed at 1:34AM Eastern Time on November 4th, 2020, which shows 136,155 votes for Joe Biden and 29,115 votes for Donald Trump
An update in Michigan listed as of 3:50AM Eastern Time on November 4th, 2020, which shows 54,497 votes for Joe Biden and 4,718 votes for Donald Trump
This report predicts what these vote updates would have looked like, had they followed the same pattern as the vast majority of the 8,950 others. We find that the extents of the respective anomalies here are more than the margin of victory in all three states — Michigan, Wisconsin, and Georgia — which collectively represent forty-two electoral votes.
Extensive mathematical detail is provided and the data and the code (for the data-curation, data transformation, plotting, and modeling) are all attached in the appendix to this document[1].
No comments:
Post a Comment